Ritik Agrawal

Experience

1. Senior Software Engineer (Frontend) · Impact Analytics

   May 2026 – now · Bengaluru

   Microfrontends with Webpack Module Federation; React performance work on complex components; structured domain knowledge for AI-assisted engineering.

2. Software Engineer II · Akridata

   Jan 2024 – Apr 2026 · Bengaluru

   Led Vision Assist (Electron) and Vision Command (web) for 20+ global manufacturers: inspection accuracy +30%, defect-monitoring adoption +25%. Architected a desktop UI handling 6,000+ image events a day over gRPC and Triton, kept responsive with IPC, worker threads and batching. Offline-first workflows with auto-recovery; D3 and ECharts image analysis; mentored peers.

3. Software Engineer I · Akridata

   May 2022 – Dec 2023 · Remote

   Data-heavy features in React, Vue, Redux and Vuex; broke a legacy monolith into modular, domain-driven pieces; cut initial load ~30% (3.2s → 2.2s) with Vite and code splitting.

4. Frontend Intern · Akridata

   Dec 2021 – May 2022 · Remote

   CRUD workflows and RBAC user management that cut onboarding issues ~40%.

Projects

• Vision Assist

  Akridata · 2024 – 26

  The Electron app inspectors use on the factory floor, reviewing images as they come off the line. It has to stay quick at thousands of image events a day, and keep working when the network doesn't.

  My part. I led the desktop app: the Electron architecture, the split between the main and render processes, and the path an image event takes from the inference service to the screen. I mentored the peers who built on top of it.

  What I decided

  • Kept the gRPC stream out of the render process and talked to the UI over IPC, so decoding an event never blocks a frame.
  • Batched incoming events and moved the heavy work to worker threads and async pipelines, so a burst arrives as steady updates instead of a freeze.
  • Made the inspection workflows offline-first with auto-recovery, so work continues through a network drop and reconciles when the connection returns.

  What came of it

  • Inspection accuracy up 30%, for 20+ global manufacturers.
  • 6,000+ image events a day, handled without the desktop UI stalling.
  • Offline-first workflows with auto-recovery on the factory floor.

  Electron · React · TypeScript · gRPC · Triton · IPC · Worker threads

• Vision Command

  Akridata · 2024 – 26

  The web dashboard the same manufacturers use to watch quality across their lines: heatmaps, zoomable image views and live metrics, in the browser.

  My part. I led the web app, and built the visualisation layer myself: the heatmaps, the bounding-box overlays on inspection images, the zoomable views, and the live metrics the quality teams watch.

  What I decided

  • Used D3 where the drawing needed exact control (heatmaps, bounding boxes over the image) and ECharts where a standard chart was already the right answer — less custom code to maintain.
  • Fed the dashboard from live updates over WebSockets, so the quality view matches the line without anyone reaching for refresh.
  • Made the image views zoomable in place, so a defect can be inspected without leaving the dashboard.

  What came of it

  • Defect-monitoring adoption up 25%.
  • Live quality views for 20+ global manufacturers.
  • Heatmaps, bounding boxes and zoomable image views in D3 and ECharts.

  React · TypeScript · D3 · ECharts · WebSockets

• Microfrontends

  Impact Analytics · 2026 – now

  The product is split into microfrontends so separate teams can deliver independently. I work on the seams: module boundaries, shared dependencies and how the pieces load together.

  My part. I work on Module Federation and the microfrontend patterns around it — where the module boundaries go, which dependencies are shared, and how each slice ships on its own. I also chase down unnecessary re-renders in the heavier React components, and structure the domain knowledge across modules so AI-assisted debugging is actually useful.

  What I decided

  • Share a dependency only when the download savings beat the runtime coupling it creates — the reasoning I wrote up on Medium.
  • Draw module boundaries along the domain, so a team can release its slice without a coordinated deploy.
  • Profile before optimising: find the re-renders that actually cost frames in complex components, rather than memoising everything.

  What came of it

  • Independent delivery for separate product teams.
  • Two write-ups on Module Federation's shared-dependency trade-offs.

  React · TypeScript · Module Federation · Webpack

Writing

• Module Federation shared: Versioning, Trade-offs, and When to Use It · 25 Sept 2026 · 5 min read
• The 7 MB Problem: Sharing Dependencies in Module Federation · 13 Sept 2026 · 5 min read